We help IT Professionals succeed at work.

DHCP Best practices including 80/20 rule

drothbart asked
Medium Priority
Last Modified: 2010-04-21
Our current DHCP environment is in need of overhaul. We have 60 scopes currently, running on one clustered server. This has been in place for several years, and is also a file server.

We intend to relocate this onto dedicated redundant servers. I've been reading several MS documents where they recommend setting up 2 servers with the same scope, set up with one server to issue 80% of the addresses and the other to issue 20%. Doesn't the 20% just run out and stop giving out addresses?

What do others here do? We have approximately 350 servers and 3000 desktops, along with several hundred other IP devices.

We will also be overhauling DNS and WINS, so if that matters I'd love to hear the suggestions. We use Active Directory integrated DNS with a single forest/domain.

Watch Question

you should have all scopes on both servers.  Then active 40% on one and 60% on teh other.  This was the will not cross scopes and if one dies, you just activate the other scopes.


You allocate 80% of the addresses to be given out to one server. The 20% needs to be given out by the other. The DHCP server setup will not be identical.

so, in C class a subnet DHCP server 1 would give out IPs - and DHCP2 would give out -

Per http://technet2.microsoft.com/windowsserver/en/library/b9b1845d-19f2-4f13-8a7b-95ca35b021981033.mspx?mfr=true

You set them with the same scopes and different exclusion ranges.


I understand, Brent, but how does that help since it seems to me that the one issuing 20% will give all it's addresses out before the one with 80%.

For example: A scope has 200 addresses. Server A gives out 1-160, server B gives out 161-200. If there are 150 clients, wouldn't server B be almost out of leases all the time? Is that what you see in reality?
It is possible that it will. The 80/20 is just something to go off of though, you can run a 60/40  or 50/50 setup or something else if this is a concern.

If the 20% runs out, the clients will just hit the other server. If the 80% server gets shut down for some reason, you just pull the exception list on the 20% server. If you pull the exception list, I would set the conflict detection to 1 though so that you dont get multiple IPs on the subnet at the time of failure.

Here are some other thoughts on this subject (although I dont like the third option - have both with same scope just doesnt look good to me).


Thanks, Brent. Good link that was new to me as well. Good food for thought.

Explore More ContentExplore courses, solutions, and other research materials related to this topic.